On Sun, 2002-09-29 at 11:22, tarvid wrote: > Much wisdom in the reponse but... > > 1) Stability is an issue but most of my problems show up in the disk > subsystem. I have had some nasty crashes which went away when I replaced the > drives.
The only time I have noticed this phenomenon is when some drives were mishandled, without proper static protection. And there were crashes. I have seen DOA's, but they've been rare. > 2) Performance is an issue but I suspect that too is in the disk subsystem. > Processing HTTP logs is the biggest single burder on my servers. That > generates tons of stuff few people read, so I think changing content of log > analysis will take care of that one. Mail is a close second but that one is > solved by judicious non-use of synced log files. Raid 0+1 would seem a good approach for this. > 3) When I rebuild a server, newer technology is better than older and larger > aggregates of RAM are better than smaller. The latter is probably the more > important of the two so I would rather use single 512MB PC2100 than two 256MB > faster SDRAMs. I agree with the SDRAM sentiment, but I do not agree that newer is always better. It always takes the linux driver contributors time to catch up to the latest chipsets. The HPT373 Raid chipset is a prime example of this. A concern because my next system board will have one; and I'd like to be running a distro that can use that IDE bus. So staying behind the front lines with a mobo that's going to run Linux is offtimes a smart move. It gives the manufacturers time to work the lemons out of the boards and the developers time to write the hardware support in. > 4) Workstations are a different matter. I am content with much less than a > game player but I do have an urge for decent multimedia - DVD , 6 channel > sound, fast ripping. The most notable performance issue is fast program > loading. That too is in the disk subsystem. If you are doing mp3 or video encoding then you need FPU speed; and the "game" benches are once again the way to go. > So much of Tom's hardware doesn't speak to me. I'd like to avoid spending a > lot of time on both hardware compatibility and software installation issues, > I have lots of things to do. Well, I don't think the speaking is the problem, it's the listening part that is. I've been using this site as a resource for a long time, because there is a need to evaluate mobos from the standpoints that I've given URL's for. The secret is that by the time these mobos pass their tests and give their performance results, you can be pretty sure that you've got a mobo that will work, especially if there are other mobos out there with the same chipset that are running Linux. Red Hat in the past has said that there are very few mobos that don't run their distro (the only one I've seen that they listed as problematic was a Supermicro), and that holds true for Mandrake, as long as you are choosy about where you get your information from. Roundup reviews are the best reviews that exist for mobos. A single mobo review that does not allow a system board to compete with other boards is *useless*. (Unless you originally located the mobo with a roundup.) True mobo evaluation does not occur in a vacuum, it occurs within the context of competition; and that means other system boards. Roundup reviews are *expensive* to conduct. Therefore they are rare. So when you do find one, like this eighteen motherboard review, it's worth your while to sit up and pay attention. The reason I do is because as the guy in charge of a computer fabrication line, it was my task to order 20 or more mobos for evaluation at a time (plus R&D them) before the prototyping and launch of a new model. There was no internet, there were no mobo reviews, and the trade rags were for locating and ordering, not judging. So we did it the hard way; and it cost big bucks, and big time. Basically I was doing roundup reviews at the time and did'nt realize it. Now we have the internet and the advantage of performance and stability reviews. To me, THD is like a free ride. So yes, I'm damn sure going to take advantage of them; and if I'm somewhat skittish for any reason, I'll fire up google and see if anybody's had direct linux experience with the board. Not hard to do. So far I have not needed to because there are very few linux incompatible boards out there and almost none of them make it to the THD roundup reviews. If anybody's inclined, they can certainly do a double check on the web. If another quality board with the same chipset as the one you are considering is running Linux, then that puts you in the middle of the ball park anyway. Which is the case with just about the entire review here. I also went thru a real bunch of sites before I took THD serious, so it's not like I haven't been looking around. > Unfortunately, what we see are reports of people having problems and others > speculating on the latest whiz bang technology. It would be much more useful > if someone maintained a database of "this worked for me". > > Jim Tarvid Back in "the day", there were many standards in the motherboard world that were yet to be set. As a result, back then we had more compatibility problems than we do now. Even back then, Linux worked better on that less-than-standard hardware than winblows did. And there were no mobo "compatibility lists." I depend alot on what users are saying about what they are using, and I don't put stock in the cpu manufacturer's "compatibility lists" either. In that there is too much chance for self-aggrandizement or political manuvering. These days, things are a little better ordered than they were. Not saying that there's no problems. But as I've stated previously, by starting with a roundup review, you are automatically using a filter wether you know it or not; a board that doesn't work will not get a) into the roundup or b) a good review. Pros and cons are listed plainly in the review. Again, if you find a well made board that has run the gauntlet of a large set of performance (and other) tests, and is also at the top of the performance food chain, then you've got a winner. When I was sysadmin for a web company, we had a ragtag set of machines in the building with completely various system boards. The entire server room had the look of somebody that had been playing hardware russian roulette. With zero analysis, without exception, Linux worked on every one of those machines. Here, I have an Abit KT7A Raid mobo that I ordered using the Dr Tom site. Everything worked perfectly. I also have a Tyan S2390A with a Via KT133 chipset; that one worked. I did not look at any list at all before I got that board, cause it was given to me. Yet still it worked. There are more examples; but... The truth is that if you have a quality mobo that has a chipset that is working with Mandrake on other systems (as is the case with the KT333 chipsets) and it has been reviewed as being solid under test, then you've got a good investment. You are correct in saying that the disk subsystem is most often the source of the problems; and that is exactly the subsystem that needs to be under scrutiny. The newer boards are more likely to have compatibility problems with the linux drivers; most pointedly with the IDE Raid buses. But not for too long. > > http://www17.tomshardware.com/mainboard/02q3/020905/kt400-16.html > > > > But not truly "conclusive". Because if later day bios upgrades are able > > to fix the memory standards that as of right now have not been > > finalized, then the KT4 Ultra may well be the best choice for the > > future. As you can see right now the KT333's are showing faster > > benches. A bios upgrade may well fix that, however. > > > > Regarding the KT3 Ultra, this pick is perhaps a little confusing. If > > you are considering a mobo with KT333 architecture, there are many more > > superior choices available. Here are the benchmarks for a roundup of 18 > > mobos that includes the KT3 Ultra: > > > > http://www17.tomshardware.com/mainboard/02q2/020509/kt333-29.html > > > > This in fact was the roundup that I was using to choose my next mobo. > > As of now tho, the memory architecture has moved onward and I will > > probably wait for Abit to release a KT400 mobo. The advantage with this > > set of mobos is that they have been out for awhile and they are better > > known personalitywise. It is very unlikely that you will see any > > problems with the top performers in this list. Speaking of performance, > > the KT3 Ultra is in the bottom half of both the opengl and directx > > benchmarks. Dis is not good. > > > > When I was planning on getting one of these, I was looking at three: > > > > 1) Epox EP-8K3A+ -- Pros: Top performer, solid brand, reputation for > > dependability > > > > Cons: Unsure of overclocking virtues compared to Abit AT7 > > > > 2) Enmic 8TTX2+ -- Pros: German manufacturer, looked solid, top > > performer; real fast and stable > > > > Cons: New company, no history, unsure of overclocking virtues compared > > to AT7 > > > > 3) Abit AT7 -- Pros: I have extensive prior experience with this > > brand and know the overclocking and stability virtues of this brand, the > > quality has never let me down. > > > > Cons: Not a top performer (unless overclocked !) hehe > > > > In which case it can theoretically exceed the default performance of any > > of the 18. Depending on CPU limitations. > > > > Let me smash a few myths right off the bat. 1) An overclocking mobo is > > typically superior stability (and quality) wise to a mobo that is > > typified as a "server board", AKA Tyan mobos. The overclockers HAVE to > > be more stable in worse situations than the much overpriced and > > underperforming "server boards". Anything marketed as a server board is > > probably going to be a money pipeline waiting for you to give your > > frogskins away to the manufacturer. Synopsis: The overclockers are > > better server boards than the server boards. > > > > 2) Since they are marketed to hardware hacker types, they are more > > loaded with features than other boards. Also, just because you buy an > > overclocker board doesn't mean that it's going to overclock the cpu. > > You have to TELL it to do that; otherwise it operates the CPU at the > > design spec for the CPU. In that case it theoretically is more rock > > solid than the other types. > > > > 3). The reason I cite directx and opengl benches is simply because > > there is no better stress test for the entire mobo and it's subsystems > > as a team than these tests. They stress the graphics, memory, and all > > other subsystems better than any business apps ever would in a million > > years. The business benches are more or less like tits on a boar hog. > > But, if you are still curious you can follow the other URL's on Dr Tom's > > site to the business benches. > > L8r, LX -- °°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°° Kernel 2.4.18-6mdk Mandrake Linux 8.2 Enlightenment 0.16.5-11mdk Evolution 1.0.2-5mdk Registered Linux User #268899 http://counter.li.org/ °°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com